Self-driving technology has long been touted as the future of mobility, promising of safer roads, reduced congestion, and an entirely new approach to public transportation. And with a market estimated to create up to $400 billion in revenue, it’s no surprise that initiatives from Tesla, Waymo, and General Motors’ Cruise have poured billions into developing autonomous vehicles.
But the dream of fully autonomous cars—vehicles that can drive anywhere, under any conditions, without human oversight—referred to as Level 5 autonomy, remains out of reach. Level 5 autonomy is often depicted as an inevitable next step, but the reality is far more complex. AVs today are confined to controlled environments and specific use cases for good reason. The dream of fully driverless cars navigating every road under every condition is, at least for now, closer to science fiction than deployable technology.
We spoke with Vineeth Reddy Vatti, a Machine Learning Engineer at Torc Robotics, a leading company in self-driving technology for commercial and enterprise applications.
Vatti, with expertise in AI-driven perception, multi-sensor fusion, and large-scale data engineering—plus a Best Autonomous Technology award from Titan Innovation—has been instrumental in refining AV capabilities. He offers a pragmatic perspective on where autonomy stands today—and why Level 4 is the real future of self-driving technology.
The Breakthroughs Driving AV Development
Over the past decade, there have been significant improvements in AI, sensor technology, and cloud computing, enabling self-driving vehicles to better interpret and react to their surroundings. Perception—the ability of AVs to reliably “see” and analyze their environment—has been a persistent challenge, but advancements in multi-sensor fusion have helped improve detection accuracy.
“Autonomous vehicles rely on many different inputs—LiDAR, cameras, traditional radar—to navigate,” Vatti explains. “The challenge is synchronizing and interpreting these inputs together. One approach we’re prioritizing is a shared bird’s-eye view space that fuses sensor data, which makes a significant difference for object detection.”
This marks a major step toward making AV perception more reliable in complex settings, leading to a 30% improvement in 3D object detection mean average precision. That means self-driving systems are getting better at identifying pedestrians and other vehicles—especially in complex urban environments.
These improvements are meaningful, but they are incremental, not revolutionary. While AI-driven models have become more adept at handling complex scenarios—such as predicting the movements of jaywalking pedestrians and lane-weaving cyclists—they still struggle with unpredictable real-world conditions that human drivers handle instinctively.
Roadblocks to Full Autonomy
Despite technological progress, self-driving vehicles remain far from achieving unrestricted autonomy. Most AVs today operate in geofenced areas, such as urban robotaxi services in select cities. Expanding beyond these controlled zones is one of the industry’s greatest challenges, particularly when considering the unpredictable nature of real-world driving environments.
“Level 5 autonomy is about navigating an unstructured, unpredictable world—essentially, the real world,” Vatti notes. “Unmarked roads, unexpected construction zones, extreme weather—these are all very common and frequent challenges. Humans make judgment calls based on context—teaching AI to replicate that level of intuition is exponentially more difficult.”
Another major barrier? Cost. Most AVs rely on expensive LiDAR sensors and high-performance GPUs, making large-scale deployment financially challenging. Tesla has been vocal about ditching LiDAR in favor of a camera-only approach, arguing that AI-driven computer vision can be just as effective.
“There’s some debate about whether pure vision-based systems can match the reliability of LiDAR, but it doesn’t have to be an either-or situation,” Vatti explains. ” Multi-sensor fusion—combining different data sources to compensate for each other’s weaknesses—remains the strongest option. When it comes to safety, redundancy is never a bad thing.”
The Business of Autonomy: Who Will Win the Race?
The AV industry’s trajectory is shaped by economic and regulatory realities. While tech giants continue to chase the dream of Level 5, the focus is shifting toward practical, scalable applications of autonomy. Freight trucking, industrial fleets, and controlled environments like airports and shipping ports are the most viable use cases in the near term.
And Vatti has firsthand experience optimizing AV infrastructure for cost efficiency. His team at Torc Robotics reduced cloud expenses by 45% by automating pipeline management, highlighting how operational overhead still leaves money on the table.
“Autonomy won’t happen overnight,” he says. “Widespread adoption, particularly outside of software-driven industries, follows a phased approach. The most immediate and practical applications will emerge in commercial sectors—such as freight trucking, industrial fleets, and controlled environments like airports—long before consumer vehicles become fully autonomous. Success in this space depends on achieving economies of scale and proving real-world viability before broader deployment.”
Investor interest in self-driving technology remains high, but the funding priorities have shifted. Rather than pouring money into fully autonomous solutions, investors are focusing on technologies that enhance safety, mapping, and sensor efficiency—practical solutions that bridge the gap between human-driven and fully automated transportation, and feed into more general AV readiness.
What’s Next for Self-Driving Cars?
The road to full autonomy is long, and while technological progress continues, the idea of Level 5 AVs seamlessly integrating into everyday life remains speculative. Over the next decade, Vatti predicts the following key developments:
- Level 4 Regulatory Clarity: Europe is likely to lead the way in defining clear safety and deployment standards for Level 4 AVs, paving the way for broader adoption.
- Fleet-Based Deployments: AVs will expand beyond pilot programs and become a core part of industrial and commercial fleets, particularly in logistics, freight, and delivery operations.
- Advancements in Perception & AI: Self-driving technology will become more robust against edge cases, such as adverse weather conditions, sensor occlusions, and dynamic environments.
- More Efficient AV Infrastructure: Innovations in compute efficiency, cloud-based AV training pipelines, and edge AI processing will make AVs more scalable and cost-effective.
However, as for fully autonomous, no-steering-wheel cars? They may be further away than they appear. The industry is moving forward, but at a measured pace, grounded by business realities, regulatory challenges, and the fundamental limitations of artificial intelligence.
According to Vatti, the future of autonomy isn’t an overnight revolution—it’s an iterative evolution, shaped by pragmatic engineering choices and real-world scalability.